Your public website remains the source
The default onboarding path starts from approved public URLs and sitemaps without requesting customer database credentials or private-network access.
GONGJAK AX TOOL / COMPANY WEBSITE AGENT
Turn your public website into
business knowledge that answers.
Company Website Agent does not require a site rebuild or internal database access. It collects and structures approved public pages, then connects a visitor's natural-language question to grounded answers and relevant content.
VISITOR QUESTION?
↓PUBLIC WEBSITEWWW
↓GROUNDED ANSWER✓
01 / DEFINITION
What is a Company Website Agent?
Company Website Agent is an installable AX tool that turns public company, service, product, project, notice, and FAQ pages into searchable knowledge. It answers with retrieved website evidence and returns the source page, related content, follow-up questions, and an appropriate inquiry path.
One public website → one verified knowledge boundaryAnswers stay inside retrieved evidence
Only retrieved website knowledge and allowed content IDs reach the answer model, with insufficient evidence designed to return no answer.
Install without rebuilding the site
The target integration connects a lightweight JavaScript plugin to an existing search input, button, or dedicated trigger.
02 / DIFFERENCE
More than a chatbot: a website discovery system
It does not replay a scripted conversation or let a general model guess. It connects current public content to the visitor's actual question.
LAYERSCRIPTED CHAT / SEARCHCOMPANY WEBSITE AGENT
KNOWLEDGE
Operators register every answer manually
Approved web content is collected, chunked, and indexed
ANSWER
A fixed response follows a keyword or scenario
A constrained answer is generated from relevant evidence
DISCOVERY
The conversation ends inside a chat window
Answers include sources, related pages, and follow-ups
UPDATE
Every content change requires an answer rewrite
Changed pages trigger targeted knowledge refreshes
03 / PRODUCT FLOW
One AX flow from collection to grounded answers
The product is not just a model. It is the complete flow that safely collects website content, publishes a verified index, and preserves visitor questions as signals for future content improvement.
- 01DISCOVER
Discover website URLs and content
Follow sitemaps, canonicals, and internal links, then approve inclusion and exclusion rules.
- 02INDEX
Build structure-aware RAG knowledge
Preserve headings and content structure, create searchable chunks, and index them for lexical and semantic retrieval.
- 03ANSWER
Connect evidence, answers, and pages
Send only sufficiently relevant evidence to the model and return answers, citations, content cards, and follow-up questions.
- 04REFRESH
Refresh only what changed
Compare content hashes and update changed documents before atomically publishing a verified index version.
04 / KNOWLEDGE SOURCE
What can become Agent knowledge?
The default scope is information a visitor can already access on the public website. Private data integrations require separate approval and security design.
Company & brand
About pages, history, organization, locations, contacts, and official policies
Services & products
Service descriptions, product details, features, audiences, and official guidance
Projects & cases
Approved projects, responsibilities, scope, and verifiable outcomes
FAQ & updates
Frequently asked questions, careers, news, notices, and frequently changing content
05 / AGENT × AEO
Company Website Agent and AEO share one knowledge foundation
AEO helps search engines and AI answer services understand a site's entities and answers. Company Website Agent connects the same public knowledge to real visitor questions. Designing both together keeps human-facing answers aligned with externally discoverable evidence.
Explore GONGJAK AEO- 01
Clear entities
Keep company names, service names, canonical URLs, and scope consistent across copy, metadata, and structured data.
- 02
Answer-first content
Publish actual visitor questions as definitions, comparisons, procedures, and accessible FAQ content.
- 03
Evidence-page links
Preserve canonical URLs and safe internal links so every Agent answer can lead back to its source.
- 04
Visible content gaps
Do not force answers when evidence is missing. Preserve those questions as signals for deciding what public content to add.
06 / INSTALLATION
How it fits into an existing website
This is not a full-site migration. We agree on the public information scope, create a verified search index, and connect an Agent entry point to the current interface.
- 01REGISTER
Register the primary website
Register the primary domain and public site, then verify ownership and allowed origins.
- 02REVIEW
Review collection and refresh policy
AI proposes URL types and fetch modes; an operator approves what is included and excluded.
- 03PUBLISH
Validate and publish the RAG index
Only a knowledge version that passes sample questions and quality checks becomes visitor-searchable.
- 04INSTALL
Connect the JavaScript plugin
Attach an installation key and trigger to an existing button or search input to open the Agent.
07 / TRUST BOUNDARY
Trust boundaries for public website search
A visitor Agent needs tenant isolation, source validation, and an explicit insufficient-evidence path—not just a convenient interface.
- 01
Search only within the customer, site, locale, and active index
CHECK - 02
Verify the allowed Origin and site installation key
CHECK - 03
Return not_found without an LLM call below the relevance gate
CHECK - 04
Revalidate every citation ID and URL returned by the model
CHECK - 05
Keep the previous live index if a new version fails quality checks
CHECK - 06
Treat collected content and visitor questions as untrusted data
CHECK
08 / FAQ
Company Website Agent FAQ
Direct answers to the questions a team asks before installing a website Agent.
What is a Company Website Agent?
It is a website AX tool that turns an existing public website into searchable knowledge and uses that evidence to answer visitor questions with relevant source pages.
How is it different from a chatbot or site search?
It does not replay a fixed script or only list title-keyword matches. It retrieves relevant website evidence and, when evidence is sufficient, returns a constrained answer, sources, content cards, and follow-up questions.
Do we need to rebuild our website?
A rebuild is not the default integration path. The platform uses the public site as the knowledge source and is designed to connect a JavaScript plugin to an existing search input, button, or dedicated trigger. Final scope depends on the site's architecture and security policy.
Does it require access to our internal database?
No customer database account, private-network access, or internal API key is required for the public-website onboarding path. Information not supported by public pages remains a content gap rather than an invented answer.
What happens when website content changes?
The target flow follows an approved refresh policy, compares page signals and content hashes, processes only changed documents and chunks, and publishes a verified index version.
Will it answer questions without evidence?
No. The intended behavior applies relevance and minimum-evidence gates and returns not_found below the threshold. Citation IDs and URLs are revalidated by the server.
How does Company Website Agent support AEO?
The Agent and AEO use the same company entities, answer-first copy, canonical URLs, and public evidence. Questions that lack evidence can also show where the website needs clearer public content.
Can I try it now?
The GONGJAK home page already provides an Agent grounded in GONGJAK's company, service, and project knowledge. The reusable platform version for external websites is in release preparation, and its final scope may change after validation.
Where can I find pricing and delivery estimates?
Pricing and fixed delivery estimates are not published on the website. We first review the number of pages, content types, refresh policy, and required safety profile before providing a scoped proposal.
COMPANY WEBSITE → BUSINESS KNOWLEDGE